In this issue of Blood, Kurosawa et al use a Markov decision process (MDP) of allogeneic hematopoietic cell transplantation (alloHCT) versus chemotherapy in patients with acute myeloid leukemia (AML) in first complete remission (CR1).1
“In probability theory, a stochastic or random process is the counterpart to a deterministic process…. Instead of dealing with only one possible reality of how the process might evolve over time (eg, solutions of an ordinary differential equation), in a stochastic or random process there is some indeterminacy in its future evolution described by probability distributions.”2 To simplify, how can one quantify complicated processes for which there is uncertainty, often at each process step or transition? Andrei Andreyevich Markov, a Russian mathematician, is the father of understanding and quantifying stochastic processes; decision processes that contain elements of uncertainty are often named after him. An online tutorial discusses partially observable MDPs.3
Several investigators have attempted to determine the optimal approach to the clinical management of AML patients in CR1. Strategies have focused on disease characteristics that would predict for a higher risk of relapse such as cytogenetic and molecular markers, ability to attain remission, and other factors including patient age, comorbidities, and donor availability. AlloHCT is associated with a higher early risk of treatment-related mortality (TRM) than chemotherapy and has a unique risk of graft-versus-host disease (GVHD). However, alloHCT is accompanied by the salutary graft-versus leukemia (GVL) effect that significantly decreases the risk of AML relapse compared with chemotherapy. How can these risks be analyzed along with their effects on quality of life (QOL)? MDP has been implemented for medical decision making when evaluating therapy options along with economic, QOL, and/or outcome analyses.4 However, it has not been used frequently for decision making with regard to the use and timing of alloHCT for hematologic disorders. The first report of MDP for AML patients in CR1 was published in 1996 when the rate of alloHCT-TRM was higher, when only younger patients were eligible for alloHCT and without the benefit of cytogenetic and molecular risk stratification.5 Furthermore, MDPs were conducted for alloHCT in the treatment of chronic myeloid leukemia in the pre- and post-imatinib eras, myelodysplastic syndrome in the pre-demethylation agent era, and also for sickle cell anemia.6-9 The analysis of posttreatment outcomes is complex with competing risks for death such as chemotherapy toxicity, GVHD (alloHCT only), infection, and AML relapse. Thus, statistical models that account for competing risks by multistate modeling or landmark analyses allow clinicians to better determine the risk of various patient outcomes for clinical recommendations.10,11
With a life-threatening disease such as AML, optimizing treatment to prolong survival is the ultimate goal. However, for many patients, in the words of Abraham Lincoln “…it is not the years in your life that count. It is the life in your years.” In the Kurosawa et al analysis, QOL for each state (no relapse without GVHD, no relapse with GVHD, relapse and death for alloHCT in CR1 vs no relapse, relapse, second remission after salvage alloHCT and relapse and death [the lowest QOL measure]) was evaluated by physicians but not by patients. Future analysis might prospectively measure patient-reported QOL outcomes in patients pre- and posttreatment to determine long-term effects of various treatments on QOL that are important to patients. Comparing patient-reported outcomes with physician-assigned QOL—and determining the complexity of interpatient variability in their valuation of QOL and satisfaction with their treatment decision-making—will be very informative. Some individuals place greater value on surviving to experience more of life's milestones, while others place greater value on the quality of that existence; this applies to physicians, patients, and their caregivers. Understanding decision-making in an era of complex treatment choices will allow physicians and patients to make more informed decisions when faced with the prospect of choosing alloHCT or chemotherapy as the best therapeutic option.
When faced with self-determination (“To do is to be”: Socrates) versus determinism (“To be is to do”: Sartre), preservation of existence (“To be or not to be”: Shakespeare) matters in the context of QOL.
Conflict-of-interest disclosure: The authors declare no competing financial interests. ■
